AI in medicine, explained for everyoneGeneral information only, not medical advice · Our standards

Cancer · Diagnosis and imaging

Swapping one of the two mammogram readers for AI found slightly more cancers

How it was checked
AI-assisted source check. Not clinically reviewed. ()
Last updated
Next review
Page ID
TR-041

General information only, not medical advice. Talk to your doctor about your situation. Read the disclaimer.

In Stockholm, 55,581 women's screening mammograms were read by two radiologists and by an AI. One radiologist plus the AI caught 261 cancers, against 250 for the usual two radiologists: about 4% more. It was one hospital, women weren't randomised, and cancers found between screens weren't counted in the main result. Lunit, the AI's maker, helped fund it. Rating: In clinical trials.

How far away is this?

A large prospective study in a real screening programme, peer-reviewed in The Lancet Digital Health (2023). Women weren't randomised to different reading set-ups, and it ran at one hospital, so it's In clinical trials, not yet proven routine care.

Where is it approved?

RegionRegulatorStatusNotes
United StatesFDAUnknownRegister not searched yet.
European UnionCE mark / EMAUnknownRegister not searched yet.
United KingdomMHRAUnknownRegister not searched yet.
AustraliaTGAUnknownRegister not searched yet.

We haven't checked which regulators have approved the AI used in this study. "Register not searched yet" means we haven't searched that regulator's register; it doesn't mean the product is or isn't available. Approval in one country isn't approval in another.

What is it?

In Sweden's screening programme, the usual standard is for every mammogram to be read by two radiologists. This study at Capio Sankt Göran Hospital asked a simple question: what if the AI replaced one of them? Every mammogram was read by everyone, so the researchers could compare the combinations on the same women.

What's the evidence?

  1. From April 2021 to June 2022, 55,581 women aged 40 to 74 having routine screening were included. 269 of them (0.5%, about 1 in 200) were diagnosed with breast cancer found at screening.[1]

    Evidence stage: In clinical trials · #c1

  2. One radiologist plus the AI found 261 cancers; two radiologists found 250. That's 11 more cancers out of 55,581 women, a relative rise of 4%, and it met the study's test for being no worse (non-inferior).[1]

    Evidence stage: In clinical trials · #c2

  3. The AI reading alone found 246 cancers versus 250 for two radiologists, and two radiologists plus the AI found 269. Both were also non-inferior to standard double reading.[1]

    Evidence stage: In clinical trials · #c3

  4. It was one hospital, and cancers were counted within 3 months of the mammogram, so cancers that show up between screens weren't part of the main result. Funding came from the Swedish Research Council, the Swedish Cancer Society, Region Stockholm and Lunit, the AI's maker. The authors say the AI has potential for controlled use, with risk management and real-world follow-up of its performance.[1]

    Evidence stage: In clinical trials · #c4

What does it mean for me?

If your screening service uses AI as one of the readers, this study suggests it can find about as many cancers as two radiologists. It ran at one hospital and didn't count cancers that appear between screens, so it isn't the final word. Keep going to your screens, and see your doctor about any breast change, even if your last mammogram was clear. It's fair to ask your service whether AI is part of reading your mammogram.

What could I ask my doctor?

  • Does my breast screening service use AI to read mammograms?
  • If AI is one of the readers, does a radiologist still look at my images?

These are general starting points, not a recommendation to choose any particular test or treatment.

If you're worried about your health right now, contact your doctor or local health advice line. In an emergency, contact your local emergency number.

Change history

  • : First published.

General information only, not medical advice. Talk to your doctor about your situation. AI-assisted source check. Not clinically reviewed.

References

  1. The Lancet Digital Health (2023): AI for breast cancer detection in screening mammography in Sweden (ScreenTrustCAD). doi.org (2023). Peer-reviewed study. https://doi.org/10.1016/S2589-7500(23)00153-X
  2. PubMed record 37690911 (abstract). pubmed.ncbi.nlm.nih.gov. Abstract. https://pubmed.ncbi.nlm.nih.gov/37690911/

Publisher shown as the website hosting each source. Dates are given where the source label includes them.